[cfe-commits] r169571 - /cfe/trunk/lib/StaticAnalyzer/Core/RegionStore.cpp

[Jordan Rose]

On Dec 7, 2012, at 10:27 , Ted Kremenek <kremenek at apple.com> wrote:

> On Dec 7, 2012, at 9:36 AM, Jordan Rose <jordan_rose at apple.com> wrote:
>>>        // invalidate that region.  This is because a block may capture
>>>        // a pointer value, but the thing pointed by that pointer may
>>>        // get invalidated.
>>> -        Store store = B.getRootWithoutRetain();
>>> +        Store store = B.asImmutableMap().getRootWithoutRetain();
>>>        SVal V = RM.getBinding(store, loc::MemRegionVal(VR));
>> 
>> This is an unfortunate impedance mismatch.
> 
> Impedance mismatch?  Please elucidate.
> 
>> 
>> 
>>> StoreRef RegionStoreManager::BindArray(Store store, const TypedValueRegion* R,
>>> @@ -1724,9 +1813,9 @@
>>> 
>>>  // There may be fewer values in the initialize list than the fields of struct.
>>>  if (FI != FE) {
>>> -    RegionBindings B = GetRegionBindings(newStore.getStore());
>>> -    B = addBinding(B, R, BindingKey::Default, svalBuilder.makeIntVal(0, false));
>>> -    newStore = StoreRef(B.getRootWithoutRetain(), *this);
>>> +    RegionBindingsRef B = getRegionBindings(newStore.getStore());
>>> +    B = B.addBinding(R, BindingKey::Default, svalBuilder.makeIntVal(0, false));
>>> +    newStore = StoreRef(B.asImmutableMap().getRootWithoutRetain(), *this);
>>>  } 
>> 
>> Same impedance mismatch in BindStruct, BindVector, and BindArray. For that matter, it's worse: we know that all the bindings will happen in a single cluster.
> 
> Not sure what you mean by impedance mismatch.  I do agree that asImmutableMap().getRootWithoutRetain() chain is gross, and can be sugared up.  Is that what you mean?

In both cases, we're converting back and forth between Store(Ref) and RegionBindingsRef. Before that was annoying but not a huge performance hit, but now it means we're canonicalizing earlier than we otherwise would.

Jordan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20121207/51548cbc/attachment.html>